Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Older Taxpayers’ Response to Taxation of Social Security Benefits
Leonard Burman, Syracuse University and Tax Policy Center Norma B. Coe, University of Washington and NBER
Kevin Pierce, Internal Revenue Service
Liu Tian, Syracuse University1
1 Some of the research reported herein was pursuant to a grant from the U.S. Social Security Administration
(SSA) funded as part of the Retirement Research Consortium (RRC) through the Boston College Center for
Retirement Research (CRR). The findings and conclusions expressed are solely those of the authors and do
not represent the views of the SSA, any agency of the federal government, the RRC, the Urban Institute,
Syracuse University or Boston College. The authors would like to thank: Zhenya Karamcheva for
programming assistance; Rachel Johnson for supplying estimates from the IRS public use file; Dan Feenberg
for help with TAXSIM; Jan Ondrich, John Sabelhaus, seminar participants at the Center for Policy Research
at the Maxwell School of Syracuse University and at the IRS - Tax Policy Center 2013 Joint Research
Conference for helpful comments.
I. Introduction Social Security benefits are taxed under a complex regime that raises marginal
effective tax rates by up to 85 percent. Over a range of Modified Adjusted Gross Income
(MAGI),2 affected taxpayers must include in their taxable income $0.50 of Social Security
benefit for every additional dollar of other taxable income;3 at higher income levels, $0.85
of benefits must be added, until 85 percent of Social Security benefits are included. In these
income ranges, an additional dollar of other taxable income increases total taxable income
by $1.50 or $1.85. At the highest income levels, this can convert a modest 25 percent
statutory tax rate into a 46.25 percent marginal rate. This is much higher than the top
income tax bracket,4 but it applies to older households with relatively modest incomes.
The tax on benefits is in some ways similar to the Social Security earnings test
(SSET), which reduces Social Security benefits by 50 cents for every dollar earned above
an exempt amount for those younger than the Full Retirement Age (FRA, currently 66).5
However, the taxation of benefits applies at all ages while the SSET applies only to Social
2 MAGI includes most of the income and adjustments reflected in adjusted gross income (AGI), but it includes
one-half of Social Security benefits, rather than the taxable portion. It also includes tax-exempt interest. 3 That is, any taxable income included in MAGI other than Social Security benefits. 4 In 2013, the top income tax bracket is 39.6 percent and applies to households with taxable incomes over
$450,000 (married) and $400,000 (single). 5 A 33-percent reduction and a higher exemption apply to workers in the year in which they reach FRA.
2
Security recipients who claim benefits before reaching FRA. Moreover, unlike the benefit
tax, the SSET is not a pure tax since the reduced current benefits translate into higher
benefits once FRA is reached. In contrast, the tax on benefits has no actuarial adjustment.
While the tax on benefits could have significant effects on behavior, it has been thus
far largely ignored in the literature. This is a potentially important oversight. If taxpayers
understand the rules, one would expect them to be even more sensitive to this work
disincentive than to the SSET, which most research has found to significantly affect labor
supply. Moreover, this tax not only affects earnings but also non-labor income, so it can
influence non-labor decisions, such as when to realize capital gains. Early retirees may be
subject to both the SSET and Social Security benefit taxation, so the effective combined
work disincentive may be quite large. Further, if the tax is inefficient, reform options might
exist that could bolster the trust fund, extend older people’s attachment to the labor force,
significantly reduce tax compliance costs for older workers, and raise overall economic
welfare.
This paper investigates older taxpayers’ response to the taxation of Social Security
benefits by looking for evidence of bunching at the kink points created by the taxation of
benefits. In theory, some individuals with incomes above the taxation thresholds have an
incentive to reduce their incomes to the threshold—by working less, delaying realization of
3
capital gains, or using other techniques to reduce reported income. We test this hypothesis
using a panel of data from individual income tax and information returns.
We find no evidence of bunching at or around the thresholds for the population as a
whole, and only a very small response for single self-employed taxpayers who have
previously been found to be more sensitive to changes in tax rates (Saez 2010; Chetty et al.
2011). This implies that this complicated tax does not lead to any important behavioral
response and therefore imposes little or no deadweight loss.
The paper continues as follows. Section II describes the taxation on Social Security
Benefits. Section III surveys the relevant literature. Section IV develops a simple
theoretical model. Section V discusses the data and Section VI presents the empirical
results. Section VII summarizes our findings and discusses planned future work.
II. Taxation of Social Security Benefits Prior to 1983, Social Security benefits were not subject to income tax. In 1983, the
Greenspan Commission recommended that a portion of benefits be subject to income
taxation, with the resulting additional tax revenue allocated to the OASDI (Old Age
Survivors and Disability Insurance, or Social Security) trust fund. Legislation enacted in
1993 increased the amount of benefits included in taxable income for higher-income
taxpayers, with the additional revenues allocated to the HI (Medicare) trust fund.
4
The formula for taxation is complex. OASDI benefits become subject to income
taxation when MAGI exceeds $25,000 for single ($32,000 for married) taxpayers. Above
those thresholds, the taxable portion of benefits phases in starting at a 50-percent rate. Fifty
cents of benefits are included in taxable income for every additional dollar of MAGI. After
a second threshold ($34,000 for singles and $44,000 for married households), the phase-in
rate increases to 85 percent. The phase-in continues until 85 percent of Social Security
benefits are included in taxable income.
The thresholds for taxation have been fixed in nominal terms since their inception.
Since the thresholds are not adjusted for inflation, they decrease in real terms over time,
unlike federal income tax brackets and many other income tax parameters. As a result,
taxation of Social Security affects an increasing proportion of beneficiaries over time,
pushing people into higher tax brackets. The number of returns with taxable Social Security
benefits nearly tripled—from 5.3 million to 15.3 million—between 1990 and 2009 (see
Figure 1). The dollar amount of Social Security benefits subject to taxation increased even
more, from $33.6 million in 1990 to $174.6 million in 2009, in part because of the 1993
legislation and partly because of increases in nominal income of the elderly.
5
Although the levy may seem to be a tax on Social Security benefits, it is actually a
large implicit surtax on all income included in MAGI.6 Taxpayers with low Social Security
benefits or modest amounts of other income have MAGI below the threshold for taxation
and are not affected. However, as either benefits or other income increase, effective
marginal tax rates may increase quite dramatically. For example, a single person with
$15,000 of non-Social Security income and $19,900 of Social Security benefits has none of
her Social Security included in taxable income; her marginal income tax rate equals the
6 Note that the tax potentially applies to taxpayers collecting disability and survivor benefits under the OASDI
program, but our analysis will focus on Social Security beneficiaries.
0
20
40
60
80
100
120
140
160
180
200
0
2
4
6
8
10
12
14
16
18
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
Taxa
ble
Bene
fits
in M
illio
ns o
f $20
09
Num
ber o
f Ret
urns
wit
h Ta
xabl
e Be
nefit
s, in
Mill
ions
Figure 1. Number of Returns with Taxable Social Security Benefits, and Amount in $2009, in Millions, 1990-2009
Amount of taxable benefits, in millions of $2009 (right axis)
Number of returns with taxable benefits, in millions
Source: IRS, "Selected Income and Tax Items for Selected Years (in Current and Constant Dollars),"http://www.irs.gov/file_source/pub/irs-soi/09intba.xls.
6
statutory rate of 10 percent. If either her Social Security benefit or income increases by
$100, her marginal tax rate would increase to 15 percent.
The taxation of Social Security benefits increases effective marginal tax rates by 50
percent in the first phase-in range and by 85 percent in the second. This is because an
additional dollar of AGI (earnings or non-labor income) increases MAGI by $1.50 in the
50-percent phase-in range and by $1.85 in the higher interval, until 85 percent of Social
Security benefits are included in taxable income. Figure 2 illustrates how the taxation of
benefits distorts effective tax rates for a taxpayer with $20,000 in Social Security benefits
in 2010. The effective tax rate schedule is marked by significant discontinuities—much
larger than under the regular income tax. Over the phase-in range of income, a taxpayer
would ordinarily face three marginal rates—10, 15, and 25 percent. However, because of
the partial inclusion of Social Security benefits, three additional effective rates are
created—22.5, 27.75, and 46.25 percent. The top effective rate, which applies to seniors
with relatively modest incomes ($33,000-$39,000 in Figure 2), is actually higher than the
top statutory income tax rate of 35 percent that applied to households with taxable income
over $373,650 in 2010.
As shown in Figure 2, taxpayers with income just beyond the phase-in region face a
marginal rate of 25 percent, which is more than 20 percentage points lower than those with
7
lower incomes. Taxation of benefits reduces their after-tax income, but there is no implicit
surtax or marginal disincentive to work or earn other income.
The implicit tax affects not only earnings but also non-labor income. Burman (1999)
points out that the taxation of Social Security can have disproportionate effects on effective
long-term capital gains tax rates; it can add up to 21.25 percentage points (85 percent of 25
percent) to the statutory capital gains tax rate of 15 percent that applies to taxpayers in that
income range.
If there is a behavioral response to the taxation of benefits, the substantial kinks in
the tax schedule could create clustering of households at the kink points, and potentially
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
0 10,000 20,000 30,000 40,000 50,000
Non-Social Security Income, in Dollars
Figure 2. Effective Marginal Tax Rates for Single non-Itemizer, Age 66 or Older, With $20,000 of Social Security, by non-Social Security
Income, 2010
50% phase-in rate
85% phase-in rate
No implicit surtax
No implicit surtax
Statutory Rate
8
discourage labor supply at both the extensive and intensive margins. Although taxpayers
with very low and very high non-labor income are likely to be unaffected, taxpayers whose
earnings would be subject to partial taxation might be less likely to work than other similar
taxpayers. Secondary earners may face especially strong disincentives if the primary
earner’s income puts the second earner in the phase-in range.
The tax treatment of benefits could also affect decisions about when to begin
claiming Social Security. The steeply rising marginal tax rate schedule creates an incentive
for many people to claim benefits early, getting a reduced benefit over more years.
Individuals born after 1942 can reduce their annual benefit by 25 percent or more by
claiming at age 62 rather than the full retirement age and fully or partially avoid taxation of
Social Security benefits. As a result, the adjustment for delayed retirement may no longer
be actuarially fair when taxes are considered. On the other hand, some taxpayers may have
an incentive to delay claiming Social Security benefits. If a worker reaches the full
retirement age and expects to keep working for a few more years after which his non-Social
Security income would drop significantly, he may elect to delay claiming Social Security
benefits if the future drop in income means that much less of his benefits would be subject
9
to tax. In this case, the after-tax value of delaying retirement is better than actuarially fair,
even if before tax, the trade-off is neutral.7
Finally, it should be noted that the very complicated taxation of Social Security
benefits might affect behavior much differently than predicted by a pure optimizing model.
It is possible that people do not understand how the tax affects marginal tax rates, the
incentives on labor supply, or the timing of benefits. If people ignore these incentives, then
the tax may be a type of optimal tax—raising revenue with little or no effect on behavior.
On the other hand, taxpayers may overreact to misunderstood incentives—magnifying the
economic distortion.
III. Previous Literature
While Social Security has been extensively studied, very little attention has been
paid to the taxation of benefits. The closest analogue is the SSET, which reduces Social
Security benefits for individuals who have not reached the full retirement age and whose
earnings exceed a threshold.8 The SSET is different in several key ways. For one thing, it
is much easier for individuals to determine if they are affected since it depends only on
7 Coile et al. (2002) model the timing of claiming Social Security. Even ignoring the taxation of social Security
benefits as they do, the decision is very complicated. They present nonlinear simulations for the case of a single
earner, leaving the more complex case of dual earners to later research. They find that men generally claim
benefits too early compared with the optimal choice. 8 Prior to 2001, there was also a SSET at a reduced rate for individuals between the full retirement age and 69.
10
individual earnings and age. In another sense, though, it is more complicated because there
is an actuarial adjustment. The reduced Social Security benefits translate into higher future
benefits (assuming the individual lives long enough to claim them) making labor supply
decisions a function not only of the tax rate, but life expectancy and discount rates.
Evidence, however, suggests that older workers view the SSET as a tax with little or no
awareness of the actuarial adjustment. Several studies find evidence that the SSET
discouraged work among older Americans. 9 Also, eliminating the earnings test for
beneficiaries who had reached the full retirement age increased the likelihood that workers
would claim Social Security benefits before age 70 (Song and Manchester 2007; Friedberg
and Webb 2009).
The Social Security benefit formula itself impacts the implicit taxes on work. The
formula is progressive, so those with high earnings get much less in additional benefits per
dollar of payroll tax than those with lower incomes. For some workers, including those who
expect to have fewer than 40 covered quarters of work—and are thus ineligible for
benefits—or who will receive benefits based on their spouse’s earnings, the payroll tax is a
pure tax. Liebman, Luttmer, and Seif (2009) find labor supply and retirement decisions of
9 Friedberg (2000), Gustman and Steinmeier (2005), Benitez-Silva and Heiland (2007), Song and Manchester
(2007), Heider and Loughran (2008), Engelhardt and Kumar (2009), and Friedberg and Webb (2009). Burtless
and Moffitt (1985), Gruber and Orszag (2003), Gustman and Steinmeier (1985) and Song and Manchester
(2007) find small effects.
11
older workers to be sensitive to the variation in the effective tax rate on earnings. Of
particular relevance, this research suggests a surprisingly sophisticated understanding of
complex rules. A survey by Leibman and Luttmer (2012) finds a fair amount of knowledge
of some Social Security provisions and relatively less about others (including the earnings
test).
We know of only three previous studies that have examined the taxation of Social
Security benefits. Goodman and Liebman (2008) look at the taxation of benefits as a form
of means-testing and conclude that it is sub-optimal. They do not explicitly consider the
effect of taxation of benefits on economic incentives, but, citing behavioral economics
research, they question whether and how individuals might respond to the tax incentives:
While this analysis shows that the taxation of Social Security benefits raises marginal tax rates
for a sizable minority of Social Security beneficiaries, the complexity of these provisions raises
questions about how future and current beneficiaries perceive these incentives and whether
their behavior responds to them. (Goodman and Liebman 2008, pp. 17-18)
One possibility is that, overwhelmed by the complexity of the incentives, taxpayers might
simply ignore the tax. Alternatively, they might apply a simple rule of thumb—e.g., on
average, 4 percent of Social Security benefits are included in income—that could similarly
result in little distortion. Or, Goodman and Liebman (2008) conjecture, taxpayers may
misperceive the tax as applying to 85 percent of Social Security benefits. This could create
12
a quite large income effect—even for taxpayers with incomes so low that little or none of
their benefits are taxable—although presumably it would have no effect on the perceived
after-tax return to working or earning other income.
Page and Conway (2011) measure the income effect of taxation of benefits directly
by exploiting the natural-experiment of introduction of the taxation in 1983, using
difference-in-differences methodology with data from the Current Population Survey (CPS).
They estimate that a 20 percent reduction in after-tax Social Security benefits boosts labor
force participation among high income elderly by 2 to 5 percentage points. They argue that
taxation of Social Security benefits increases labor supply through the income effect:
people above the threshold where 85 percent of benefits are subject to tax, even before
including OASDI benefits, have less after-tax income, which increases hours of work. They
do not attempt to measure the marginal effect of reduced after-tax income within the
phase-in range.
Burman, Coe, and Tian (2011) attempt to measure directly the effect of taxing
Social Security benefits on labor force participation and earnings using data from the
Health and Retirement Study (HRS). They do not find evidence that taxation of benefits
significantly affects labor market behavior, but they raise the major caveat that their
estimates may be unreliable because of errors in variables and small sample size. Survey
13
estimates of tax information are notoriously imprecise and the HRS lacks key components
of taxable income, such as capital gains.
IV. Effects of Taxing Social Security Benefits
If taxpayers understand how the taxation of Social Security benefits affects their
budget, then we should observe bunching of MAGI near the thresholds. Taxing Social
Security benefits generates convex kinks in the budget constraint at the thresholds for the
50-percent and 85-percent phase-in rates (corresponding to MAGI of $25,000 and $34,000
for single filers). In a simple model of utility maximization, taxpayers with incomes only
slightly greater than the threshold will reduce their incomes to the threshold.
To see this, consider a simplified example in which there is a flat-rate income tax
and only one rate of taxation of Social Security (as was the case between 1983 and 1993),
which increases tax rates by 50 percent. The optimal level of MAGI will maximize utility
subject to the kinked budget constraint (Figure 3). Assuming that individuals are averse to
work and other activities that increase MAGI and that they value consumption (after-tax
income), higher utility corresponds to indifference curves that move in a northwesterly
direction on the figure.
Figure 3 illustrates three categories of taxpayers who will be affected differently by
the introduction of taxation of benefits. In Panel A, MAGI in the absence of taxation of
14
Social Security would fall below the threshold. That individual is unaffected by benefit
taxation. Panel B shows a taxpayer who before the tax change would have MAGI of
z*+Δz*, but after introduction of the taxation regime chooses MAGI of z*. Saez (2010)
shows that in the case where individuals have identical preferences but differ in their ability
to earn income (e.g., their hourly wage rate differs), all individuals with initial incomes
between z* and z*+Δz* would bunch at the kink. Taxpayers who initially have higher
incomes than z*+Δz* may also reduce their incomes, but their new incomes would be
tangent to the new budget constraint to the right of z*. Finally, Panel C depicts high-income
taxpayers for whom the tax produces only an income effect.
Figure 3. Effect of Introducing a Kink in the Budget Constraint
15
With perfect information and complete ability to choose MAGI, this framework
would produce bunching at the threshold z* (see Figure 4). The kink has no effect on
16
taxpayers with initial incomes below z*, but it produces a leftward shift in the distribution
of income among those with initial incomes above z*. Saez (2010) extends this analysis to
allow for adjustment frictions (e.g., people can only imperfectly adjust income or they have
imperfect information about the location of the threshold) and shows that under certain
simplifying assumptions, the amount of bunching near z* provides a measure of the
compensated elasticity of taxable income. If individuals are very sensitive to taxation (high
elasticity), then there will be an unusually large mass of tax returns near the threshold.
Figure 4. Illustration of Bunching at Threshold (z*) in Simple Utility Maximization Framework
There are many contexts in which such bunching may be observed. Saez (2010)
shows that self-employed individuals’ incomes tended to bunch at the level where the
earned income tax credit starts to phase out. Wage earners showed no such response, which
is consistent with the notion that the self-employed have more control over hours worked
Den
sity
Before-Tax Income
z* z*+Δz*
before tax density
aftertax density
pre-policy incomes between z* and z*+Δz* bunch at z* after tax change
17
and taxable earnings, and self-employment income is not subject to third-party information
reporting, making it easier to misreport on a tax return. Friedberg (2000), Song and
Manchester (2007), Engelhardt and Kumar (2009), etc. observe that older workers clustered
to the left of the SSET exempt threshold. Chetty, et al. (2011) examine bunching around
large jumps in tax brackets in Denmark to measure elasticity of taxable income in the
context of search costs.
Our hypothesis is that if taxpayers are aware of the incentives created by the
taxation of Social Security benefits, there should be a bump in the empirical distribution of
tax returns near the two thresholds for taxation. We would expect the bump to be more
pronounced for those with income from self-employment.
V. Data
To look for evidence of bunching, we use administrative data—the 1999 IRS Statistics
of Income (SOI) Individual Edited Panel, which is a longitudinal dataset drawn from
individual income tax returns and information returns.10 The data comprise a panel of
individuals from tax years 1999 to 2008. The advantage of these data is that they provide an
accurate measure of what is reported to the IRS on income tax returns—and thus tax status.
10 For more information on the SOI Individual Income Tax Return Panel, see Weber and Bryant (2005).
Pierce (2011) documents an extended version of the panel (through 2008).
18
They also allow us to study the behavior of self-employed individuals; those who previous
research suggests would be the most responsive. The disadvantage is that the dataset
includes little demographic information, which precludes structural modeling of the
response to taxation.
The panel has been augmented by matching all of the primary and secondary SSNs
within the panel to the SOI-processed information returns databases for Forms W-2
(information on wages and withholdings), Forms 5498 (contributions to retirement
accounts), Forms 1099-SSA (Social Security benefits), and Forms 1099-R (income from
retirement accounts and pensions). Separate observations are created for primary and
secondary taxpayers who were in the sample in 1999. The panel is a stratified random
sample, which oversamples high-income returns. Sampling weights allow estimation of
population aggregates.
We use information from several tax forms for the analysis. Our measure of gross
Social Security benefit comes from Form 1099-SSA, an information return the Social
Security Administration produces to report benefits for each recipient. Tax-exempt interest
and the amount of Social Security benefits that are included in AGI come from Form 1040.
Our dataset also includes reported self-employment income from Form 1040 Schedule SE.
19
Administrative data are not immune from measurement error. For example,
self-employed taxpayers often misreport their income to the IRS.11 The data, however,
accurately reflect pre-audit information that taxpayers report to the tax authorities, and the
resulting tax liability. Therefore, any behavioral response to the taxation of Social Security
benefits should be evident on the tax return.
The sample starts with 112,823 records in 1999, but diminishes to 106,655 by 2008
(see Table 1). We are primarily interested in the subsample of taxpayers age 62 and over,
which includes 23,535 individuals in 1999 and 36,530 in 2008. The weighted sample
includes 153.6 million individuals in 1999, 28.6 million of whom are age 62 and over.
Attrition within the panel is primarily due to death, but taxpayers may also drop out in
years in which their income falls below the filing threshold. Because our sample has been
supplemented with information returns, particularly earnings from the W-2 and Social
Security benefit payments from Form 1099-SSA, we will continue to observe almost all
individuals who are not required to file an income tax return. The sample of
interest—taxpayers age 62 and over—actually increases over time, a reflection of an aging
sample population.
11 Based on audit data, only 43 percent of nonfarm proprietor income (i.e., small business income) was
voluntarily reported on tax returns in 2001 (Internal Revenue Service 2006).
20
Table 1. 1999 SOI Edited Panel Sample Sizes, 1999-2008
Tax Year Total Sample Subsample with Primary
Taxpayer Age 62 or Over
Unweighted Weighted Unweighted Weighted
1999 112,823 153,578,941 23,535 28,574,758
2000 112,804 153,904,818 24,797 29,666,609
2001 112,783 154,187,313 25,990 30,637,473
2002 112,528 154,118,136 27,282 31,720,317
2003 112,058 153,648,715 28,536 32,640,930
2004 111,144 152,282,996 30,269 34,106,473
2005 110,048 150,512,455 31,918 35,426,133
2006 108,946 148,771,365 33,380 36,666,559
2007 107,844 147,034,343 34,740 37,831,748
2008 106,655 145,134,423 36,530 39,309,668
Note: Total sample excludes returns receiving disability payments and those where the primary taxpayer is younger than 23.
All told, the panel includes 755,087 observations for married individuals and
352,546 for singles, representing multiple annual observations for most individuals (see
Table 2). Applying sample weights, that represents 906.9 million married filers and 606.3
million single filers. Most of the sample is too young to qualify for Social Security benefits
(see Table 1); only 21.4 percent of married individuals and 18.6 percent of singles have
Social Security benefits.12
Table 2. Summary Statistics, Pooled Sample: 1999-2008
Married Single
12 Younger adults may qualify for Social Security disability benefits, but those individuals have been excluded
from our sample.
21
Unweighted Weighted Unweighted Weighted
Number of Returns 755,087 906.9M 352,546 606.3M
Self-employed (%) 28 19 9.6 7.3
With SSB income (%) 23.5 21.4 20.1 18.6
Social Security Benefit ($) 24,371 20,489 15,171 13,711
SSB in AGI ($) 14,802 8,605 5,415 3,036
MAGI ($) 3,146,938 110,823 1,050,723 37,518
wage earners
Wage income ($) 508,710 75,881 112,727 27,040
Self-employed
Self-employment income ($) 428,950 37,070 341,050 20,111
Wage income ($) 1,085,714 65,144 646,428 17,189
VI. Results
Figure 5 reports the distribution of MAGI relative to the first exempt amount calculated
using the IRS Panel. A value of -1,000 on the x-axis means $1,000 below the threshold.
Most of the panels are restricted to the sample of taxpayers who have been claiming Social
Security benefits for at least 1 year under the logic that it may take time to understand the
tax rules. Results are very similar if that restriction is lifted, and also are similar at the
second threshold for taxation (see Appendix). Relative MAGI is measured in 2008 dollars.
All of the histograms are weighted by population weights; unweighted histograms (not
shown) look similar.
Figure 5. MAGI distribution around the first threshold
22
To examine bunching evidence statistically, we compare the empirical density,
represented by the dots in the scatter plot with smoothed distributions, indicated by the
solid line, of MAGI in the vicinity of the threshold in the right panel of each figure. The
smoothed distribution is fitted by a quadratic form of MAGI, excluding the observations
within $1,000 of the threshold. The grey band indicates the 95 percent confidence interval,
reflecting the underlying variability of the data. The simple empirical test of bunching is
whether observations near the threshold fall outside the confidence band (reflecting normal
sample variability).
23
Unlike the histograms for the SSET, EITC, or Danish tax system reported in earlier
studies, there is no visual evidence of bunching near the MAGI threshold, indicated by the
red line, either for all taxpayers or for the self-employed subsample.
It is possible that married and single taxpayers respond differently to the taxation of
benefits. Single taxpayers have an easier optimization problem to solve so this is a cleaner
test of the bunching hypothesis. Presumably singles have more control of their own MAGI
than individual spouses have in managing joint MAGI. Figure 6 shows the MAGI
distribution separately for married and single households. Although there is no evidence of
bunching for wage earners, there is a hint of bunching to the left of the threshold for single
taxpayers with income from self-employment.
All told, the evidence would seem to allay concerns that taxpayers might be
over-reacting to the taxation of Social Security benefits. Responses appear to be modest, at
most. There is only weak evidence of response for single taxpayers with self-employment
income.
24
Figure 6. MAGI distribution around the first threshold by marital status
VII. Conclusion
The taxation of Social Security benefits creates high effective marginal tax rates,
which gives older workers an incentive to reduce their labor and non-labor income below
the taxable threshold. However, the tax rules are also quite complex. While in theory
taxpayers have an unambiguous incentive to reduce income in the neighborhood of the
threshold, the practical effect of these complex incentives is an empirical question. If
taxpayers respond to those incentives, there could be significant efficiency costs as well as
implications for Social Security’s and the nation’s finances as older workers would be
25
paying less income and payroll taxes. Moreover, the issue is important as the nation
considers tax reform options, which might include changing the way Social Security is
taxed.
This study uses administrative data from tax and information returns to examine the
distribution of Social Security recipients in the neighborhood of the taxation thresholds.
There is little evidence of a response. We examined married and single individuals with and
without self-employment income. Only single, self-employed people show any evidence of
reducing income to avoid the tax and the response is much smaller and less precisely
estimated than the response Saez (2010) found to the kink in the EITC benefit schedule.
Overall, the findings suggest that older taxpayers have little understanding of the incentive
effects of taxing Social Security.
In future work, we plan to look at how taxation affects labor force participation and
the timing of capital gains realizations; capital gains face a much larger proportional rise in
tax rates than other income and the timing of capital gains realization is comparatively easy
to manipulate. We also plan to look at whether the taxation of benefits affects when
individuals first claim Social Security benefits.
26
VIII. References
Benitez-Silva, Hugo and Frank Heiland. 2007. “The Social Security Earnings Test and
Work Incentives.” Journal of Policy Analysis and Management 26(3): 527-555.
Burtless, Gary and Robert A. Moffitt. 1985. “The Joint Choice of Retirement Age and Post
Retirement Hours of Work,” Journal of Labor Economics. 3:209-236.
Burman, Leonard E. 1999. The Labyrinth of Capital Gains Taxation: A Guide for the
Perplexed. Washington, DC: Brookings Institution Press.
Burman, Leonard, Norma Coe, and Liu Tian. 2011. “Do Older Taxpayers Respond to the
Taxation of Social Security Benefits?” Manuscript.
Chetty, Raj, Chetty, Raj, John N. Friedman, Tore Olsen, and Luigi Pistaferri. 2011.
“Adjustment Costs, Firm Responses, and Micro vs. Macro Labor Supply Elasticities:
Evidence from Danish Tax Records.” The Quarterly Journal of Economics 126(2):
749-804.
Coile, Courtney, Peter Diamond, Jonathan Gruber, and Alain Jousten. 2002. “Delays in
Claiming Social Security Benefits,” Journal of Public Economics, 84(3): 357–385.
Engelhardt, Gary V. and Anil Kumar. 2009. “The Repeal of the Retirement Earnings Test
and the Labor Supply of Older Men.” Journal of Pension Economics and Finance
8(4): 429-50.
Friedberg, Leora. 2000. “The Labor Supply Effects of the Social Security Earnings Test.”
The Review of Economics and Statistics 82(1): 48-63.
Friedberg, Leora and Anthony Webb. 2009. “New Evidence on the Labor Supply Effects of
the Social Security Earnings Test.” Tax Policy and the Economy 23: 1-35.
Engelhardt, Gary and Kumar, Anil. 2009. “The repeal of the retirement earnings test and
the labor supply of older men.” Journal of Pension Economics and Finance.
Song, Jae G. and Manchester, Joyce. 2007. “New evidence on earnings and benefit claims
following changes in the retirement earnings test in 2000.” Journal of Public
Economics 91 (3-4): 669-700.
Gruber, Jonathan and Peter Orszag. 2003. “Does the Social Security Earnings Test Affect
Labor Supply and Benefits Receipt?” National Tax Journal 56(4):755-773
27
Gustman, Alan L. and Thomas L. Steinmeier. 1985. “The 1983 Social Security Reforms
and Labor Supply Adjustments of Older Individuals in the Long Run.” Journal of
Labor Economics 3(2): 237-253.
Gustman, Alan L. and Thomas L. Steinmeier. 2005. “The Social Security Early Retirement
Age In A Structural Model of Retirement and Wealth.” Journal of Public Economics
89(2-3): 441-463.
Heider, Steven J. and David S. Loughran. 2008. “The Effect of the Social Security Earnings Test on Male Labor Supply: New Evidence from Survey and Administrative Data.”
Journal of Human Resources 43(1): 57-87.
Internal Revenue Service. 2006. “IRS Updates Tax Gap Estimates.” Available at
http://www.irs.gov/uac/IRS-Updates-Tax-Gap-Estimates.
Liebman, Jeffrey. 1997. “The Impact of the EITC on Incentives and Income Distribution.”
In Tax Policy and the Economy, ed. James Poterba, 83-119. Cambridge: MIT Press.
Liebman, Jeffrey and Serena Goodman. 2008. “The Taxation of Social Security Benefits as
an Approach to Means Testing.” Working Paper NB08-02. Cambridge, MA:
National Bureau of Economic Research.
Liebman, Jeffrey B., Erzo F.P. Luttmer, and David G. Seif. 2009. “Labor Supply Responses
to Marginal Social Security Benefits: Evidence from Discontinuities.” Journal of
Public Economics 93(11-12): 1208-1223.
Liebman, Jeffrey B., and Erzo F. P. Luttmer. 2012. “The Perception of Social Security
Incentives for Labor Supply and Retirement: The Median Voter Knows More Than
You'd Think,” in Jeffrey Brown, ed., Tax Policy and the Economy, Volume 26.
Cambridge, MA: National Bureau of Economic Research.
Page, Timothy, Conway, Karen. 2011. “The Labor Supply Effects of Taxing Social Security
Benefits.” University of New Hampshire (unpublished manuscript).
Pierce, Kevin. 2011. “General Description Booklet for the SOI Edited Panel Retirement
File.” Internal Revenue Service (unpublished manuscript).
Saez, Emmanuel. 2010. “Do Taxpayers Bunch at Kink Points?” American Economic
Journal: Economic Policy 2 (August 2010): 180-212
Song, Jae and Joyce Manchester. 2007. “New Evidence on Earnings and Benefit Claims
Following Changes in the Retirement Earnings Test in 2000.” Journal of Public
Economics 91(3-4): 669-700.
28
Weber, Michael, and Victoria Bryant. 2005. “The 1999 Individual Income Tax Return
Edited Panel,” 2005 Proceedings of the American Statistical Association, Social
Statistics Section, Government Statistics Section, American Statistical Association.
29
Appendix. Graphical Examination of Bunching Around the Second Taxation
Threshold
This appendix shows graphs of the empirical density of tax returns around the
second (higher) threshold at which Social Security benefits are phased into taxable income
at an 85 percent rate (increasing marginal effective tax rates by 85 percent), There is no
significant evidence of bunching around this threshold for Social Security recipients.
Figure A1. MAGI distribution around the second threshold
30
Figure A2. MAGI distribution around the second threshold by marital status